VoIP Voice and Fax Signal Processing

4

GENERIC VAD/CNG FOR WAVEFORM CODECS

The voice activity detection (VAD) algorithm classifies the input signal as active speech or silence. The purpose of VAD is to reduce the number of packets during inactive speech and to send all available packets during active speech. Comfort noise generation (CNG) recreates a suitable background as a continuity of the inactive region. For complete end-to-end operation, VAD is required on the sending side and CNG is required on the receiving side. VAD is for detecting and giving a decision between speech and silence. Silence insertion description (SID) is the payload name or silence packet created from the sender. The packet transmission from the sender side is controlled by a discontinuous transmission (DTX) algorithm, which resides on the VAD operation/module. For sending a SID packet from the sender, information from both the VAD and the DTX controls are involved. The next transmission of SID packets will happen during a change of power or spectral characteristics. In actual usage, VAD, SID payload formulation, and DTX functions are referred to by the single name of VAD. The name “VAD packet” is commonly used for representing the “SID packet.” In this book, the VAD packet is also used at several places as this quickly conveys a VAD operation, but the correct name is “SID packet.” The combination of transmitter and receiver is referred to as VAD/CNG. CNG is for comfort noise creation at the receiver based on received SID packets. CNG uses ...